Gene network and pathway analysis of bovine mammary tissue challenged with Streptococcus uberis reveals induction of cell proliferation and inhibition of PPARγ signaling as potential mechanism for the negative relationships between immune response and lipid metabolism

General considerations

The microarray analysis clearly indicated that the mammary gland after a 20 h inoculation with S. uberis experienced a wide transcriptional response, which encompassed > 2,000 genes. Overall, the functional analysis uncovered that few functions were significantly affected, i.e. immune response was clearly the most affected and induced followed by cell proliferation/cycle/death and transport of protein and ions both of which were induced. In contrast, lipid metabolism was inhibited. Cell proliferation was seemingly or evidently induced in IPA analysis but GO analysis revealed that regulation of cell proliferation was not clearly towards induction, suggesting that the process of proliferation was probably increased for certain cells (e.g., immune and endothelial cells) but was not important for others. The overall picture from IPA and GO analyses captured the most affected functions but did not provide information of the potential mechanisms at play. The use of well-established pathways (i.e., canonical pathways) together with information about single genes provided additional means to unravel the mechanisms controlling the mammary response to IMI before peak clinical signs.

Top up-regulated DEG

Numerous cytokines involved in the immune response were significantly up-regulated in mammary tissue during IMI challenge with S. uberis. Genes coding for the cytokines TNF, IL6, and IL1B were among the top DEG (Table 1). Furthermore, among the list of all DEG (i.e., 2,102) there were several up-regulated DEG that belong to pro-inflammatory pathways including CD14, TRAF6, NFKBIA, NFKB2, and STAT3 (Figure 1). Functional analysis with GO and IPA clearly underscored the induction of inflammation as well as cytokine binding in mammary tissue from IMI (Figure 1 and Additional File 15). Other cytokines or cytokine receptors up-regulated during S. uberis IMI included IL18 and IL1R2 (verified via qPCR; Table S1). Interleukin-18 (IL-18) can induce interferon gamma (IFN-γ) production from T cells and, in combination with IL-12, IL-18 can inhibit IL-4-dependent immunoglobulin (Ig) E and IgG1 production as well as activate IgG2a secretion by B cells [25]. However, microarray analysis indicated that expression of IL12 and IFNG was not significantly altered during IMI challenge with S. uberis and that IL4R expression increased (1.26-fold; Additional File 2). This may indicate that the up-regulation of IL18 had minimal downstream affects on the innate immune response to S. uberis. Similar results were observed by Yang et al. [26] where IL18 expression, but not IFNG, was up-regulated in MEC after IMI challenge with Staphylococcus aureus (Staph. aureus).

The chemokine IL8 had the greatest change in expression resulting in a fold-change of 1,054 in infected vs. control quarters (Table S1). The importance of this protein and its related functions was underscored by GO molecular function analysis both when the entire DEG or those with >1.5-fold were analyzed (Additional Files 2 and 15). This chemokine is induced upon stimulation of TNF or IL-1 (Figure 1) and serves as a primary chemoattractant during the innate immune response, thus, playing a major role in the chemotaxis of PMN. Therefore, it seems logical that the dramatic increase in IL8 expression would occur before peak clinical signs of mastitis. Swanson et al. [8] did not observe a significant change in IL8 expression in bovine mammary tissue after IMI with S. uberis (Strain 233); but increased IL8 mRNA expression has been reported in primary isolates of bovine MEC after challenge with Escherichia coli (E. coli) [27]. With regards to results of Swanson et al. [8], mammary tissue was collected between 60-132 hours post-inoculation when peak clinical signs already had occurred. In our study, mammary biopsies were performed prior to peak clinical response and prior to the major influx of PMN into the mammary gland (supported by gene markers analysis, Figure S2 in Additional File 1), milk compositional changes and clinical signs of mastitis [14], and our previous work using this dose and strain of S. uberis [17].

The anti-inflammatory IL-10 and pro-inflammatory IL-6 pathways are activated before peak clinical signs

Interleukin-10 Signaling was among the primary canonical pathways affected by IMI challenge with S. uberis (Table 3 and 4, Figure 1, and Additional Files 7 and 12). The binding of the IL-1 cytokine family to the IL-1 receptor mediates the activity of TRAF6 (tumor necrosis receptor-associated factor 6; Figure 1 and Additional File 12), leading to activation of the p38 MAPK signaling pathway that ultimately leads to increased transcription of IL10. Despite a significant down-regulation of p38 MAPK (i.e. MAPK12; -1.22-fold; Figure 1) during IMI, the observed 13.9-fold up-regulation of IL1B and 38.9-fold up-regulation of IL1R2 probably overcame that response and also might have been sufficient to account for the fact that JAK1 expression was down-regulated (-1.17-fold; Figure 1). Interleukin-10 is an anti-inflammatory cytokine that blocks NF-κB activity, which leads to suppression of pro-inflammatory mediators such as TNF, IL6, and IL1. Expression of 22 out of 25 putative components (71 total in IPA) of the IL-10 signaling pathway present in our microarray platform were moderately but significantly up-regulated (Figure 1). Of interest was the mild up-regulation of STAT3 (ca. 3-fold; Table S1) which in turn is known to activate SOCS3 and activate IL-6-signaling [28]. Despite the marked up-regulation of IL10, our results of pathway analysis were indicative of more pronounced inflammation and probably hampered IL-10 anti-inflammatory activity.

Interleukin-6 Signaling was a major pathway affected by IMI challenge with S. uberis (Tables 3 and 4 and Figure 1). Several genes overlap between IL-6 and IL-10 Signaling, including an up-regulation of TNF, IL1B, NFKBIA, STAT3, TRAF6, and FOS and down-regulation of MAPK12. Expression of IL6 occurs via the NF-κB signaling pathway. Interleukin-6 is a pro-inflammatory cytokine that is also involved in acute-phase protein signaling (Figure 2). The coordinated up-regulation of genes involved in both IL-6 and IL-10 Signaling during IMI with S. uberis is suggestive of an ability of the immune system to generate a pro-inflammatory response via the IL-6 Signaling Pathway while attempting to control the severity and duration of the inflammation via the anti-inflammatory IL-10 Signaling Pathway. By far, however, our data suggested that the pro-inflammatory response via IL6 and TNF overrode the anti-inflammatory response via IL10.

IL-6 also has been shown to have anti-inflammatory capabilities through inhibition of IL-1β and TNF [29–31]. In our study, however, the signaling pathway through TNF and IL-1 appeared largely activated (Figure 1), with a more pronounced up-regulation of NFKBIA (in the pathways in Figure 1, IκB expression is determined by this gene, which is one of its components) than NFKB2 (this genes is a component of NF-κB), which suggested a potential inhibitory effect on the induction of survival genes via NFKB2 [32] and a control of inflammatory response. In the context of regulation of cell death/survival, it was evident that cell survival via enhanced growth and differentiation might have been inhibited due to IMI because the genes coding for phosphorylation enzymes in the ERK (extracellular-regulated kinase)/MAPK (mitogen activated protein kinase) signaling pathway, which is involved in control of a broad range of intracellular functions [33] were down-regulated (Table 3). These data suggested that signaling through phosphorylation (see also PI3K/AKT signaling; Table 3, Additional Files 7 and 12) was inhibited as a result of IMI. This latter finding was also observed by GO analysis (Additional File 15).

Our results regarding IL-6 and IL-10 support the work of Swanson et al. [8] who observed an up-regulation of the IL6 receptor (1.83-fold change) and IL10 receptor alpha (1.91-fold change) in bovine mammary tissue after S. uberis IMI. Similar to our data, Lutzow et al. [7] after IMI challenge with Staph. aureus observed an up-regulation (via microarrays) of genes involved in both IL-6 and IL-10 signaling including IL1B, IL6, IL8, CD14, FOS, and NFKBIA. In our study, we isolated whey from all infected quarters and analyzed samples for concentrations of IL-10, IL-1β, and TNF at 0, 12, and 20 h (time of biopsy) post-inoculation [14]. No significant changes in cytokine concentrations were observed by 20 h post-inoculation when compared to pre-infection levels (0 h). This may be attributed to the fact that the mammary biopsies were performed prior to peak clinical signs of mastitis in order to avoid tissue samples with elevated amounts of mRNA from infiltrating PMN. Unfortunately, the side effects of the biopsy procedure (e.g., blood clots) made it impossible to isolate whey from mammary secretions during peak clinical signs. However, Bannerman et al. [15] evaluated cytokine secretions in whey collected from mammary quarters challenged with S. uberis and observed elevated milk concentrations of IL-1β, IL-8, IL-10, IL-12, TNF, and IFN-γ compared with healthy quarters by ~30 h post-challenge.

Toll-like receptor signaling

Bacteria contain pathogen-associated molecular patterns (PAMPs) motifs, such as LPS or lipoteichoic acid (LTA), that are potent stimulators of innate immunity. Lipid A is considered the active motif for the PAMP activity of LPS from Gram-negative bacteria such as E. coli that stimulates the innate immune response and activates TLR-4 and the LPS-LPS binding protein-CD14 complex; however, the active motif for the PAMP activity of LTA (i.e. Gram-positive bacteria such as S. uberis and Staph. aureus) remains unknown. Regardless, TLR-2 protein is activated via LTA. The toll-like receptor (TLR) signaling pathway results in the synthesis of several pro-inflammatory cytokines (TNF, IL1B, and IL6) and chemokines (IL8). Although this pathway was not among the most significant in IPA analysis (Table 3 and Additional Files 6, 7, 8, 9 and 12), several genes involved in TLR signaling were up-regulated during IMI challenge including TLR2, TLR4, CASP8, CD14, FOS, IRAK1, TRAF6, and NFKBIA. All genes were verified via qPCR (Table S1). In addition, TOLLIP (toll interacting protein), a negative regulator of inflammation, was also significantly up-regulated (1.15-fold change; Additional File 2).

Several studies have evaluated gene expression profiles in mammary tissue or MEC lines after challenge with another Gram-positive bacterium, Staph. aureus [6, 7]. Lutzow et al. [7] observed that Staph. aureus alters both TLR-2 and TLR-4 signaling pathways. They observed an up-regulation of TLR2, FOS, and NFKBIA during IMI challenge with Staph. aureus as well as TLR4 and CD14, both of which are primarily activated via LPS from Gram-negative bacteria such as E. coli. These researchers also observed an up-regulation of pro-inflammatory mediators including TNF, IL1B, IL8, and IL6 after IMI with Staph. aureus. However, Yang et al. [26] observed that IMI challenge with high doses of Staph. aureus (10,000 cfu; Strain 1027) failed to activate NF-κB signaling and the pro-inflammatory genes TNF and CXCL8. A "weak" immune response may be attributed to the virulence factors associated with this strain of Staph. aureus, because heat-inactivated Staph. aureus increased the expression of TLR signaling components and NF-κB activation [26]. The TLR-mediated NF-κB activation not only signals numerous pro-inflammatory genes but also other anti-microbial immune defense genes such as beta-defensins, which are oxygen-independent peptides that have potent anti-microbial activities [34]. Our data also showed a significant increase in expression of BNBD5, the most abundantly-expressed member of the beta-defensin family of bactericidal peptides in MEC (4.19-fold change; Table S1; Additional File 2) [6]. Our data support results from Swanson et al. [35], who found increased expression of lingual antimicrobial peptide (LAP), a member of the beta-defensin family, during IMI challenge with S. uberis.

Our microarray analysis demonstrated an increased expression of both TLR2 and TLR4 after IMI challenge with S. uberis compared with control quarters. However, Swanson et al. [8] observed an up-regulation of TLR2 but not TLR4 expression in bovine mammary tissue after S. uberis IMI. Increased expression of both TLR2 and TLR4 signaling pathways during IMI challenge with Gram-positive or Gram-negative bacteria has been observed in recent studies [7, 26, 36]. Most of these studies have examined TLR expression patterns in response to E. coli or Staph. aureus, both major pathogens associated with mastitis in the dairy industry. Goldammer et al. [6] reported an increased expression of both TLR2 and TLR4 (8-to-12-fold change) in bovine mammary quarters naturally infected with S. aureus when compared to healthy quarters. This response is supported by results of Yang et al. [26], where both TLR2 and TLR4 were up-regulated after IMI challenge with either Staph. aureus or E. coli. Similar results were also observed when bovine MEC were challenged with LPS [36], as well as in mammary tissue after IMI challenge with Staph. aureus (determined via microarrays) [7].

Other DEG involved with immune response

Other DEG of interest that were significantly up-regulated during IMI challenge with S. uberis included HLA-DRA (1.82-fold change; Table S1 in Additional File 1; Additional File 2) and C1QC (1.37-fold change; Table S1 in Additional File 1; Additional File 2). HLA-DRA codes for the major histocompatability complex type II (MHC II) DR alpha and is primarily expressed on T lymphocytes and macrophages. This gene is considered a candidate gene marker of disease resistance [37]. The role of MHC II in mammary tissue is unclear. Fitzpatrick et al. [38]observed expression of MHC II-positive cells in the connective tissue of the healthy mammary quarters and quarters infected with formalin-killed S. uberis; although individual cell identification was not conducted. Swanson et al. [8] reported an up-regulation of HLA-DRA (1.73-fold change) in bovine mammary tissue after S. uberis IMI. The MHC II complex presents antigen fragments to T-helper cells by binding to the CD4 receptor on T-helper cells. However, we did not detect differential expression of CD4. Although mammary tissue was thoroughly blotted with gauze to remove any visual milk secretions, it is possible that the expression of HLA-DRA may have been acquired through milk lymphocytes and macrophages present in mammary tissue during the biopsy. MHC II expression in MEC warrants further investigation. The observed up-regulation of C1QC was opposite to results from both Swanson et al. [8] who found down-regulation (-1.74-fold change) after S. uberis IMI and those of Günther et al. [27] who observed a 1.6 to 3.2-fold decrease in mRNA expression of factors associated with the C1 complex (e.g., C1qA, C1qB, C1s and C1r) in bovine MEC after challenge with E. coli. The complement component C1q is the first step in the initiation of the classical pathway of the complement cascade [39]. Researchers have not been able to quantify C1q concentrations in mastitic milk and primarily attribute this to its large size (900 kDa), which may render it impermeable to the mammary epithelium [40].

No current information is available on the use of the lectin pathway in the mammary gland during an IMI. Researchers have concluded that the mammary gland must lack the classical pathway and therefore must rely primarily on the alternative pathway of the complement system [39]. The initial step of the alternative pathway involves the cleavage of complement component 3 (C3) into fragments C3a and C3b. The expression of C3 (via microarray and qPCR) was significantly up-regulated (1.43-fold change) in mammary quarters infected with S. uberis and supports the work of Swanson et al. [8]. The C3 component has been quantified in mastitic milk [41]. C3 is also a downstream intermediate step involved with both the classical and alternative pathways of the complement system that ultimately leads to the assembly of the membrane attack complex (MAC complex), which consists of complement proteins C5a, C6, C7, C8, and C9. The membrane attack complex plays a role in the disruption of the bacteria cell walls during the immune response.

Two genes involved in inhibition of the complement cascade were significantly up-regulated in infected versus non-infected mammary quarters. These genes were CD59 (1.22-fold change) and CD55 (2.07-fold change) (Additional File 2). CD59 is involved in the inhibition of the assembly of the membrane attack complex. CD55, or the decay accelerating factor for complement, binds to both the C2-C4b complex of the classical pathway and the C3-Cfb complex of the alternative pathway. This binding accelerates their decay, disrupting the cleavage of C3 into C3b and C3a fragments, which leads to inhibition of the cascade and prevention of damage to host cells. To our knowledge, this is the first report of a significant up-regulation in expression of the C1QC gene from mammary quarters infected with S. uberis. Swanson et al. [8] observed an inverse relationship between C1Q expression (-1.74-fold change) and C3 (2.36-fold change) after S. uberis IMI. The researchers did not elaborate on the inverse relationship in gene expression patterns between C3 and C1Q. Further research related to the classical pathway of the complement cascade in the mammary gland is needed.

Cell proliferation, angiogenesis, and apoptosis

The overall functional analysis both in IPA and GO clearly indicated an induction of proliferation of several types of cells but in particular immune, endothelial, and muscle cells. In contrast, several significantly-enriched pathways related to proliferation/angiogenesis were strongly (e.g., IGF1 in Figure 3B and ephrin receptor in Table 3) or likely inhibited (Aryl Hydrocarbon Receptor signaling; Additional Files 7 and 9), with both the platelet-derived growth factor (PDGF) and PI3K/AKT signaling pathways likely induced (see below).

Integration of lipid metabolism and inflammation: possible role of LXR/RXR and PPAR signaling pathways

Both LXRs and PPARs are involved in the regulation of metabolic and inflammatory signaling [49, 50]. PPARA is expressed in liver, brown adipose tissue, heart, and muscle tissue and plays a pivotal role in fatty acid catabolism [49]; whereas, PPAR-γ (PPARG) is highly expressed in adipose tissue and macrophages and primarily regulates adipogenesis [50, 51]. PPAR-γ has been shown to be expressed in bovine mammary tissue and is also significantly increased during lactation [51]. PPARA and PPARG have anti-inflammatory properties [50, 52]. PPARG has been shown to interfere with the transcription of pro-inflammatory factors such as STAT and NF-κB in macrophages [53].

In non-ruminant macrophages, studies have shown that ligand-activated LXR inhibits expression of genes involved with immune response [54]. Interestingly, studies have also shown that TLR4 activation in macrophages inhibits LXR signaling [55]. Activation of inflammatory signaling pathways and release of inflammatory mediators are fundamental to the diverse immune functions of macrophages, and the mammary gland possesses resident macrophages [56]. In addition to inducing genes involved in reverse cholesterol transport, LXR reciprocally represses a set of inflammatory genes after bacterial lipopolysacharide (LPS), TNF, or IL-1β stimulation [57]. Examples of such genes include those involved in generation of bioactive molecules such as NOS2A, IL-6, TNF, and IL-1β, the chemokines CCL2, and matrix metallopeptidases. We found that IMI resulted in marked up-regulation of IL6 (430-fold), TNF (45-fold), IL1B (14-fold), and CCL2 (3.3-fold) and moderate but significant up-regulation of NOS2A (1.2-fold) and MMP7 (1.4-fold; Table S1 and Additional File 2). As previously stated, most of the responses in the present study are likely attributed to MEC and potentially resident macrophages, which constitute ca. 5% or more of the parenchyma tissue [18]. It is possible that increased NOS2A expression may be attributed to resident macrophages. However, studies have reported increased expression of the endothelial (eNOS) and inducible (iNOS) forms of nitric oxide synthase in human [58] and murine [59] breast cancer tissue. The increased TLR4 expression after IMI in our study may partly explain the down-regulation of the genes involved with LXR/RXR signaling. The TLR4 response might have been driven via up-regulation of IRF6 (Additional File 2) [57].

Studies investigating the LXR/RXR signaling pathway in the mammary gland are sparse and have primarily focused on expression of genes involved in this pathway during murine lactation regardless of bacteriological status [60]. Mouse mammary microarray data [60] has suggested the potential involvement of two systems in controlling fatty acid metabolism. These include the LXR/RXR pathway controlling 1) β-oxidation of fatty acids via LXR (also known as NR1H2)/PPAR dimers; and 2) fatty acid synthesis involving the LXR/RXR dimer, which induce expression of the sterol regulatory element-binding proteins 1 (SREBF1) and 2 (SREBF2).

The lactating bovine mammary gland does not seem to oxidize long-chain fatty acids as a source of energy [61], thus, any involvement of LXR in bovine mammary tissue might be at the level of fatty acid synthesis and/or inflammation (as in non-ruminant macrophages) [57]. However, the expression of LXR in bovine mammary tissue only increased slightly during lactation relative to pregnancy and it was not among DEG (M. Bionaz, S. L. Rodriguez-Zas, R. E. Everts, H. A. Lewin, and J. J. Loor, University of Illinois, Urbana, unpublished results). Those responses coupled with the lack of change in LXR expression due to IMI were suggestive of a minor role for LXR in mediating anti-inflammatory or lipogenic mechanisms in bovine mammary tissue.

Expression of PPARA is barely detectable in bovine mammary tissue (M. Bionaz, S. L. Rodriguez-Zas, R. E. Everts, H. A. Lewin, and J. J. Loor, unpublished results) and tends to decrease during lactation, which points to a minor role of this nuclear receptor in bovine mammary lipid metabolism. We recently showed that mRNA expression of PPARG was consistently up-regulated during lactation, suggesting that it could play a role in milk fat synthesis [51]. A role of PPARG in regulating bovine milk fat synthesis machinery was supported by recent results we obtained where treatment of MacT cells (bovine mammary epithelial immortalized cells) with rosiglitazone, a specific PPARγ agonist, resulted in coordinated up-regulation of genes involved in FA import (e.g., CD36), de novo FA synthesis (e.g., ACACA, FASN, SREBF1), and TAG synthesis (e.g., LPIN1, SCD) [62]. More importantly in the context of the present study, a recent study with PPARγ-knockout mice indicated that its absence increased utilization of long-chain fatty acids for synthesis of inflammatory lipids due to reduced TAG synthesis [63]. PPARG-knockout mice had a sustained increase in 12-lipoxygenase (i.e., ALOX5AP) activity from parturition through the end of lactation. Although we did not observe a significant effect of IMI on PPARG expression, up-regulation of ALOX5AP (ca. 6-fold; Table S1) might have been associated with increased synthesis of eicosanoids which are classical effectors of an inflammatory response. In addition, activation of PPARγ by specific agonists reduced synthesis of inflammatory cytokines in mammary epithelial cells, suggesting this nuclear receptor has an anti-inflammatory role in mammary tissue [64]. A 39-fold increase of ALOX5AP in mammary quarters challenged with E. coli in a recent study provides further support to the inflammatory role of ALOX5AP during an IMI [27].

Taken together, the above observations coupled with the down-regulation of PPARγ target genes point to PPARγ as a major player. The expression of this nuclear receptor appeared not to be affected by IMI (at the least from microarray data) but its activity probably was decreased as suggested by down-regulation of its known target genes. Similar to PPARα (Figure 4), the increase in NFκB activity might have inhibited PPARγ activity. Interestingly, insulin-induced gene 1 (INSIG1), which is involved in the inhibition of SREBP cleavage (i.e., inactivation of SREBP), and appears to be a PPARγ target gene in bovine mammary epithelial cells [62], was significantly up-regulated (1.5-fold change; Table S1). These data suggested that INSIG1 is not only under control of PPARγ but likely contributed to reduced milk fat synthesis through blockage of SREBP1 cleavage, i.e. both SREBF1 and SREBF2 are moderately up-regulated during lactation in bovine mammary tissue and could be involved in lipid synthesis through activation of acetyl-coenzyme A carboxylase alpha (ACACA) and fatty acid synthase (FASN) [51]. Unfortunately, the IPA Knowledge Base does not contain specific PPARγ pathways, thus precluding a definitive conclusion about the pivotal role of PPARγ. It is important to note that a possible role of PPARα cannot be excluded because specific PPARα co-activators or up-stream factors were down-regulated (Figure 4).

An enzyme linked to the LXR/RXR and PPARG pathways via SREBP1 in non-ruminant liver and adipose is stearoyl-CoA desaturase (SCD), which plays an essential role in TAG synthesis by catalyzing the synthesis of oleic acid via desaturation of stearic acid [65]. Oleic acid serves as a primary substrate for fatty acid binding protein 4 (FABP4) [66], and previous work in our laboratories proposed that FABP3 provides stearic acid, and other substrates, to SCD, which then provides oleic acid for FABP4 [51]. Expression of both FABP3 and FABP4 was down-regulated in infected versus control mammary quarters (-1.46 and -1.55-fold change, respectively). Expression of SCD was also inhibited in S. uberis-infected quarters (-1.64-fold change). Impaired PPARγ signaling might have been associated with the down-regulation of these lipogenic enzymes, either through down-regulation of SREBF1 or directly through decreased binding to response elements (e.g., SCD and FABP4).

Our findings highlighted a potential relationship between PPAR and LXR, two master regulators of lipid metabolism and inflammatory responses in non-ruminants [57]. The relationships between those two nuclear receptors with inflammatory conditions appear to be in two directions, i.e. their expression/activity is decreased by inflammation in mouse liver [67] and kidney [68], and an increase in their activity/expression leads to an anti-inflammatory effect [57]. Overall, our results indicated that IMI with S. uberis inhibited activity of LXR/RXR and PPAR signaling during IMI, suggesting that the anti-inflammatory effect of those pathways was not at play. We suggest that PPARγ signaling plays a primary role in mammary tissue but the activity of this nuclear receptor was probably reduced. The overall repression of lipogenic genes in S. uberis infected mammary quarters and the mechanisms involved in LXR/RXR or PPAR signaling and the fatty acid switch in the mammary gland during IMI challenge have not been elucidated and require further investigation. PPARγ has a pivotal role in immune cells as well, increasing their ability to face infections [69]. A possible role of PPARγ activation in reducing inflammation in mammary gland tissue has been previously suggested based on in vitro data [64] and our results support such a view.

Ceramides, inflammation, and lipid metabolism

Ceramide, which is involved in cell signaling, cell cycle, and regulation of protein transport from ER to Golgi, is one of the most studied sphingolipids in nature [70]. Other sphingolipids with signaling roles include sphingosine (Sph) and sphingosine-1-phosphate (S1P), which can activate NFKBIA and a cascade of inflammatory genes (Figure 1; Additional File 2) [71]. Although minor compared with TAG, sphingolipids are the third most important lipid component in bovine milk fat [72]. Formation of the milk fat globule membrane relies on sphingolipid and cholesterol availability, thus, coordinated synthesis of both compounds is pivotal to milk lipid droplet formation/secretion. Mammary tissue synthesizes sphingolipids de novo [72] from palmitoyl-CoA, leading to ceramide formation and incorporation into sphingomyelin. Thus, palmitic acid used for ceramide synthesis in mammary appears a required step and also might represent a regulatory point for FA synthesis because ceramides can inhibit this process by blocking the activity of AKT/PKB [73].

Our data revealed that ceramide signaling was markedly down-regulated (Table 3) potentially through the action of TNF (Table S1, Additional File 2). Based on the observed downregulation of lipogenic genes (e.g., ACACA, FASN; Table S1, Additional File 2) as well as serine palmitoyl transferase (SPTLC2; Additional File 2) it was apparent that ceramide synthesis was decreased, which likely explains the down-regulation of other genes that are part of its signaling pathway (Additional Files 7 and 9). The details of the pathway indicated a reduction of ceramide synthesis from sphingomyelin through activity of neutral sphingomyelinases sphingomyelin phosphodiesterase. In addition, the decrease in expression of genes involved in long-chain fatty acids import (e.g., CD36, LPL) and de novo fatty acid synthesis (e.g., ACACA and FASN) had probably reduce the amount of available palmitate for synthesis of ceramide. From our combined results, production of ceramide did not seem to be induced by pro-inflammatory state during IMI, but probably decrease. In addition, we observed an overall inhibition/decrease of ceramide downstream signaling, which clearly indicated that during IMI ceramide is not involved in apoptosis.

Significance of the immune response and milk fat synthesis

The negative relationship between DEG involved with immune response and milk fat synthesis may serve several beneficial purposes for the immune system within the mammary gland. First, the ability of phagocytes such as PMN and macrophages to engulf invading microorganisms is lower in milk when compared to PMN and macrophages that originate from the bloodstream. Milk phagocytes engulf milk fat globules instead of invading pathogens, resulting in a loss of pseudopodia needed for phagocytic capability [56]. Therefore, the less milk fat synthesized during an IMI the more likely that milk phagocytes will engulf invading bacteria instead of milk fat globules. As previously stated, S. uberis strain O140J has been shown to be more resistant to PMN phagocytosis and more capable of establishing infection when compared to a noncapsular strain [12, 13]. Decreased expression of genes involved in Lipid Metabolism (using IPA Knowledge database) has also been recently reported after IMI challenge with E. coli [27]; and suggests that reduced lipid synthesis in the mammary gland may not be pathogen specific. In addition, microarray and qPCR analyses revealed a down-regulation of LALBA (-1.46-fold; Table S1), the rate limiting enzyme in lactose synthesis, which confirmed previous findings [74]. This may indicate that, at the time of biopsy (20 h post-inoculation), lactose synthesis was reduced as suggested by previously-reported milk whey analysis of mastitic cows [75]. A decrease in lactose synthesis might help the immune system by reducing substrate (i.e. lactose) for bacteria and also preventing a potential inhibition of PMN phagocytosis by lactose [76]. Inflammation reduces protein synthesis in muscle [77], but our transcript profiling did not indicate alterations in protein synthesis in infected compared with non-infected contralateral mammary quarters. However, there was an increase in expression of CSN3 (Table S1). Furthermore, the GO analysis uncovered an evident induction of transcription, post-translational modification, transport, and localization of proteins (Additional File 15). Those findings seemed to indicate that protein synthesis in milk should not have been decreased, but the large increase in transcription and protein metabolism was probably more related to increase synthesis and secretion of inflammatory-related proteins such as cytokines or acute-phase proteins. Unfortunately, quarter milk composition was not analyzed during the infection period; therefore, changes in milk fat, protein, and lactose could not be evaluated.

Genes positively-associated with TNF

Not surprisingly, TNF was positively associated with pro-inflammatory mediators such as IL8, IL1B, and NFKBIA. The positive association with the anti-inflammatory cytokine IL10 further supports the co-regulatory mechanisms responsible for controlling the severity of the inflammatory response during an IMI [93–95].

The network in Figure 5 also shows a positive effect of TNF on the acute-phase protein SAA3. This supports the protein-level response observed in milk secretions from cows during IMI challenge with S. uberis, where milk SAA concentrations were elevated at 20 h post-inoculation when compared to pre-inoculation concentrations [14]. Serum amyloid proteins have immunological properties and the SAA3 isoform (i.e. M-SAA3) has been shown to be highly expressed in bovine MEC during mastitis [96]. Expression of mRNA for SAA3 in MEC is significantly enhanced in quarters challenged with LPS from E. coli or with Staph. aureus when compared to healthy quarters, indicating that the main source of SAA in milk during infection may be from MEC and not hepatocytes [4]. This premise is further supported by results of Eckersall et al. [97], who demonstrated that expression of M-SAA3 mRNA and haptoglobin (HP) mRNA were up-regulated during an experimental challenge with Staph. aureus and that mRNA for M-SAA3 was greater than that for HP. This increased expression of SAA3 and HP is specific to infected quarters because several studies have indicated that expression is minimal or not detectable in MEC from healthy quarters [98, 99]. It is challenging to be able to distinguish between 2 gene isoforms with a 70-bp oligonucleotide on a microarray platform. The latest annotation of our microarray identified this oligo as both SAA1 and SAA3 and it clearly depends on the tissue type (i.e. liver or mammary) as to which isoform is primarily expressed. Upon verification, we confirmed that the sequencing results of primers were specific for SAA3 (Tables S3 and S4 in Additional File 1). IPA network analysis indicated that TNF-α protein has been shown to increase SAA3 mRNA expression in mouse granulosa cells [100] and in 3T3-L1 adipocyte cell lines [101]. SAA is primarily involved in the acute phase response and has been shown to increase leukocyte adhesion [102], but no relationships between SAA3 and genes encoding SELL and selectin-P (SELP) have been identified (Figure 5). However, TNF has been shown to affect both the expression and protein release of SELP, but not SELL, in murine endothelial cells (Additional File 15) [103].

Gene network analysis also shows that TNF has a positive relationship with PLAU and PLAUR. The enzyme PLAU is required for the normal repair of wounds originating on skin [104] and, as stated earlier, S. uberis can activate the conversion of PLAU to plasmin [23]. Plasmin increases during mastitis and hydrolyzes α_s-casein, β-casein, and β-casein [105]. An increase in expression of κ -casein (CSN3; 1.82-fold change) was observed in S. uberis infected quarters (Table S1). Concentrations of κ-casein and plasmin were not quantified in milk secretions from infected quarters for this study, thus further research will be needed to investigate their correlations between mRNA expression and protein concentrations in milk as well as their specificity to S. uberis-associated mastitis.

In the nucleus, both FOS and BCL3 expression are stimulated by TNF (Figure 5). Expression of FOS was enhanced in human omental microvascular endothelial cells when incubated with TNF-α for 10 min [106], whereas BCL3, a nuclear protein primarily found in B lymphocytes, increased when human hepatocellular carcinoma cell lines (HepG2) were stimulated with TNF-α [107]. Another positive association within the network involved LTF, which competes for iron with invading microorganisms that require it for growth [108]. Watanabe et al. [109] observed a significant increase in LTF 4 h after intramammary infusion of recombinant bovine TNF-α.

Increased CD14 and TLR2 expression was observed in S. uberis-infected quarters when compared to healthy (control) quarters (Table S1). Hermoso et al. [110] observed an increase in TLR2 mRNA expression in carcinomic human alveolar basal epithelial cells (A549 cells) after stimulation with recombinant human TNF-α. Regarding CD14, TNF-α protein increased CD14 expression in rat Kupffer cells [111]. The CD14 molecule is primarily activated via the PAMP sequence associated with Gram-negative bacteria (LPS) [112], but it has been shown to increase in Gram-positive associated IMI [15].

Genes negatively-associated with TNF

As discussed above, several genes involved in milk fat synthesis were down-regulated in S. uberis-infected quarters. Network analysis by IPA indicated that the products of CD36, GPAM, FABP4, LPIN1, LPL, and SCD, known to be involved in milk fat synthesis [51], were negatively-associated with the expression of TNF. It has been demonstrated that TNF reduces expression of LPL in rat adipocytes [113]; GPAM in mouse adipocytes [114], and CD36, FABP4, and SCD in adipocytes from human, mouse, or rat [115–117]. Most researchers examining gene expression responses after IMI challenge have primarily focused on genes involved with the immune response, and very few studies [8] have examined large-scale gene expression profiles in the mammary gland during an IMI challenge with S. uberis. This research provides evidence of a role for TNF in modulation of milk fat synthesis in the mammary gland during an IMI.

Microarrays

Oligonucleotides that were flagged with "-100" by GenePix were removed from the analysis, and the remaining data were normalized to control for dye effects using the median of control elements on the microarray. In a subsequent normalization step, the log₂ normalized ratio of mammary versus reference (i.e., RNA mixture of different tissues including mammary) signal intensities were adjusted for global dye and microarray effects and normalized by Lowess. A mixed-effects model was then fitted to the adjusted ratios (mammary/reference) using Proc MIXED [119]. The model consisted of treatment (TRT; NEB and PEB), infection (INF; YES and NO), and the TRT × INF interaction. YES identifies mammary quarters inoculated with 5,000 cfu of S. uberis; and NO identifies contralateral rear control quarters (i.e., non-inoculated). The fixed effect was dye with cow and microarray as random effects. Statistical significance probability values for TRT, INF and TRT × INF effects were adjusted for the number of comparisons using Benjamini and Hochberg's FDR. For this paper, the effect of INF, regardless of TRT [14], will be discussed. Differentially expressed genes were based on FDR P-value ≤ 0.06 which corresponded to an unadjusted P ≤ 0.01. Fold change was presented as the backtransformed LSMeans (i.e., adjusted log₂ normalized ratios of mammary versus reference) of the infected versus non-infected quarters.

qPCR

After normalization with internal control genes (ICG, see above), data were analyzed using the MIXED procedure of SAS with a random effect of pair within block (day of inoculation). Class variables included cow, TRT, INF, pair, and block. The model included TRT, INF and the TRT × INF. Statistical differences were declared as significant and highly significant at P < 0.05 and P < 0.01. Trends towards significance are discussed at P < 0.10. Relative expression values are presented as least square means (LSM). The fold change was presented as the LSMeans of the infected versus non-infected quarters.